Support SO_BINDTODEVICE in ICMP sockets

Adds support for the SO_BINDTODEVICE socket option in ICMP sockets with an
accompanying packetimpact test to exercise use of this socket option.

Adds a unit test to exercise the NIC selection logic introduced by this change.
The remaining unit tests for ICMP sockets need to be added in a subsequent CL.
See https://gvisor.dev/issues/5623 for the list of remaining unit tests.

Adds a "timeout" field to PacketimpactTestInfo, necessary due to the long
runtime of the newly added packetimpact test.

Fixes #5678
Fixes #4896
Updates #5623
Updates #5681
Updates #5763
Updates #5956
Updates #5966
Updates #5967

PiperOrigin-RevId: 376271581

7 files changed, 873 insertions, 363 deletions

diff --git a/test/packetimpact/runner/defs.bzl b/test/packetimpact/runner/defs.bzl
index afe73a69a..19f6cc0e0 100644
--- a/test/packetimpact/runner/defs.bzl
+++ b/test/packetimpact/runner/defs.bzl
@@ -115,7 +115,7 @@ def packetimpact_netstack_test(
         **kwargs
     )
 
-def packetimpact_go_test(name, expect_native_failure = False, expect_netstack_failure = False, num_duts = 1):
+def packetimpact_go_test(name, expect_native_failure = False, expect_netstack_failure = False, num_duts = 1, **kwargs):
     """Add packetimpact tests written in go.
 
     Args:
@@ -123,6 +123,7 @@ def packetimpact_go_test(name, expect_native_failure = False, expect_netstack_fa
         expect_native_failure: the test must fail natively
         expect_netstack_failure: the test must fail for Netstack
         num_duts: how many DUTs are needed for the test
+        **kwargs: all the other args, forwarded to packetimpact_native_test and packetimpact_netstack_test
     """
     testbench_binary = name + "_test"
     packetimpact_native_test(
@@ -130,12 +131,14 @@ def packetimpact_go_test(name, expect_native_failure = False, expect_netstack_fa
         expect_failure = expect_native_failure,
         num_duts = num_duts,
         testbench_binary = testbench_binary,
+        **kwargs
     )
     packetimpact_netstack_test(
         name = name,
         expect_failure = expect_netstack_failure,
         num_duts = num_duts,
         testbench_binary = testbench_binary,
+        **kwargs
     )
 
 def packetimpact_testbench(name, size = "small", pure = True, **kwargs):
@@ -163,6 +166,7 @@ PacketimpactTestInfo = provider(
     doc = "Provide information for packetimpact tests",
     fields = [
         "name",
+        "timeout",
         "expect_netstack_failure",
         "num_duts",
     ],
@@ -271,9 +275,6 @@ ALL_TESTS = [
         num_duts = 3,
     ),
     PacketimpactTestInfo(
-        name = "udp_send_recv_dgram",
-    ),
-    PacketimpactTestInfo(
         name = "tcp_linger",
     ),
     PacketimpactTestInfo(
@@ -289,6 +290,10 @@ ALL_TESTS = [
     PacketimpactTestInfo(
         name = "tcp_fin_retransmission",
     ),
+    PacketimpactTestInfo(
+        name = "generic_dgram_socket_send_recv",
+        timeout = "long",
+    ),
 ]
 
 def validate_all_tests():
diff --git a/test/packetimpact/testbench/connections.go b/test/packetimpact/testbench/connections.go
index 8ad9040ff..ed56f9ac7 100644
--- a/test/packetimpact/testbench/connections.go
+++ b/test/packetimpact/testbench/connections.go
@@ -594,32 +594,50 @@ func (conn *Connection) Expect(t *testing.T, layer Layer, timeout time.Duration)
 func (conn *Connection) ExpectFrame(t *testing.T, layers Layers, timeout time.Duration) (Layers, error) {
 	t.Helper()
 
-	deadline := time.Now().Add(timeout)
+	frames, ok := conn.ListenForFrame(t, layers, timeout)
+	if ok {
+		return frames[len(frames)-1], nil
+	}
+	if len(frames) == 0 {
+		return nil, fmt.Errorf("got no frames matching %s during %s", layers, timeout)
+	}
+
 	var errs error
+	for _, got := range frames {
+		want := conn.incoming(layers)
+		if err := want.merge(layers); err != nil {
+			errs = multierr.Combine(errs, err)
+		} else {
+			errs = multierr.Combine(errs, &layersError{got: got, want: want})
+		}
+	}
+	return nil, fmt.Errorf("got frames:\n%w want %s during %s", errs, layers, timeout)
+}
+
+// ListenForFrame captures all frames until a frame matches the provided Layers,
+// or until the timeout specified. Returns all captured frames, including the
+// matched frame, and true if the desired frame was found.
+func (conn *Connection) ListenForFrame(t *testing.T, layers Layers, timeout time.Duration) ([]Layers, bool) {
+	t.Helper()
+
+	deadline := time.Now().Add(timeout)
+	var frames []Layers
 	for {
-		var gotLayers Layers
+		var got Layers
 		if timeout := time.Until(deadline); timeout > 0 {
-			gotLayers = conn.recvFrame(t, timeout)
+			got = conn.recvFrame(t, timeout)
 		}
-		if gotLayers == nil {
-			if errs == nil {
-				return nil, fmt.Errorf("got no frames matching %s during %s", layers, timeout)
-			}
-			return nil, fmt.Errorf("got frames:\n%w want %s during %s", errs, layers, timeout)
+		if got == nil {
+			return frames, false
 		}
-		if conn.match(layers, gotLayers) {
+		frames = append(frames, got)
+		if conn.match(layers, got) {
 			for i, s := range conn.layerStates {
-				if err := s.received(gotLayers[i]); err != nil {
+				if err := s.received(got[i]); err != nil {
 					t.Fatalf("failed to update test connection's layer states based on received frame: %s", err)
 				}
 			}
-			return gotLayers, nil
-		}
-		want := conn.incoming(layers)
-		if err := want.merge(layers); err != nil {
-			errs = multierr.Combine(errs, err)
-		} else {
-			errs = multierr.Combine(errs, &layersError{got: gotLayers, want: want})
+			return frames, true
 		}
 	}
 }
@@ -1025,6 +1043,8 @@ func (conn *UDPIPv4) SendIP(t *testing.T, ip IPv4, udp UDP, additionalLayers ...
 
 // SendFrame sends a frame on the wire and updates the state of all layers.
 func (conn *UDPIPv4) SendFrame(t *testing.T, overrideLayers Layers, additionalLayers ...Layer) {
+	t.Helper()
+
 	conn.send(t, overrideLayers, additionalLayers...)
 }
 
diff --git a/test/packetimpact/testbench/layers.go b/test/packetimpact/testbench/layers.go
index ef8b63db4..078f500e4 100644
--- a/test/packetimpact/testbench/layers.go
+++ b/test/packetimpact/testbench/layers.go
@@ -824,6 +824,7 @@ type ICMPv6 struct {
 	Type     *header.ICMPv6Type
 	Code     *header.ICMPv6Code
 	Checksum *uint16
+	Ident    *uint16 // Only in Echo Request/Reply.
 	Pointer  *uint32 // Only in Parameter Problem.
 	Payload  []byte
 }
@@ -849,6 +850,10 @@ func (l *ICMPv6) ToBytes() ([]byte, error) {
 	}
 	typ := h.Type()
 	switch typ {
+	case header.ICMPv6EchoRequest, header.ICMPv6EchoReply:
+		if l.Ident != nil {
+			h.SetIdent(*l.Ident)
+		}
 	case header.ICMPv6ParamProblem:
 		if l.Pointer != nil {
 			h.SetTypeSpecific(*l.Pointer)
@@ -899,6 +904,8 @@ func parseICMPv6(b []byte) (Layer, layerParser) {
 		Payload:  h.Payload(),
 	}
 	switch msgType {
+	case header.ICMPv6EchoRequest, header.ICMPv6EchoReply:
+		icmpv6.Ident = Uint16(h.Ident())
 	case header.ICMPv6ParamProblem:
 		icmpv6.Pointer = Uint32(h.TypeSpecific())
 	}
diff --git a/test/packetimpact/testbench/testbench.go b/test/packetimpact/testbench/testbench.go
index 37d02365a..d0efbcc08 100644
--- a/test/packetimpact/testbench/testbench.go
+++ b/test/packetimpact/testbench/testbench.go
@@ -57,11 +57,21 @@ type DUTUname struct {
 	OperatingSystem string
 }
 
-// IsLinux returns true if we are running natively on Linux.
+// IsLinux returns true if the DUT is running Linux.
 func (n *DUTUname) IsLinux() bool {
 	return Native && n.OperatingSystem == "GNU/Linux"
 }
 
+// IsGvisor returns true if the DUT is running gVisor.
+func (*DUTUname) IsGvisor() bool {
+	return !Native
+}
+
+// IsFuchsia returns true if the DUT is running Fuchsia.
+func (n *DUTUname) IsFuchsia() bool {
+	return Native && n.OperatingSystem == "Fuchsia"
+}
+
 // DUTTestNet describes the test network setup on dut and how the testbench
 // should connect with an existing DUT.
 type DUTTestNet struct {
@@ -99,6 +109,16 @@ type DUTTestNet struct {
 	POSIXServerPort uint16
 }
 
+// SubnetBroadcast returns the test network's subnet broadcast address.
+func (n *DUTTestNet) SubnetBroadcast() net.IP {
+	addr := append([]byte(nil), n.RemoteIPv4...)
+	mask := net.CIDRMask(n.IPv4PrefixLength, net.IPv4len*8)
+	for i := range addr {
+		addr[i] |= ^mask[i]
+	}
+	return addr
+}
+
 // registerFlags defines flags and associates them with the package-level
 // exported variables above. It should be called by tests in their init
 // functions.
diff --git a/test/packetimpact/tests/BUILD b/test/packetimpact/tests/BUILD
index b1d280f98..37b59b1d9 100644
--- a/test/packetimpact/tests/BUILD
+++ b/test/packetimpact/tests/BUILD
@@ -306,18 +306,6 @@ packetimpact_testbench(
 )
 
 packetimpact_testbench(
-    name = "udp_send_recv_dgram",
-    srcs = ["udp_send_recv_dgram_test.go"],
-    deps = [
-        "//pkg/tcpip",
-        "//pkg/tcpip/header",
-        "//test/packetimpact/testbench",
-        "@com_github_google_go_cmp//cmp:go_default_library",
-        "@org_golang_x_sys//unix:go_default_library",
-    ],
-)
-
-packetimpact_testbench(
     name = "tcp_linger",
     srcs = ["tcp_linger_test.go"],
     deps = [
@@ -410,10 +398,23 @@ packetimpact_testbench(
     ],
 )
 
+packetimpact_testbench(
+    name = "generic_dgram_socket_send_recv",
+    srcs = ["generic_dgram_socket_send_recv_test.go"],
+    deps = [
+        "//pkg/tcpip",
+        "//pkg/tcpip/header",
+        "//test/packetimpact/testbench",
+        "@com_github_google_go_cmp//cmp:go_default_library",
+        "@org_golang_x_sys//unix:go_default_library",
+    ],
+)
+
 validate_all_tests()
 
 [packetimpact_go_test(
     name = t.name,
+    timeout = t.timeout if hasattr(t, "timeout") else "moderate",
     expect_netstack_failure = hasattr(t, "expect_netstack_failure"),
     num_duts = t.num_duts if hasattr(t, "num_duts") else 1,
 ) for t in ALL_TESTS]
diff --git a/test/packetimpact/tests/generic_dgram_socket_send_recv_test.go b/test/packetimpact/tests/generic_dgram_socket_send_recv_test.go
new file mode 100644
index 000000000..00e0f7690
--- /dev/null
+++ b/test/packetimpact/tests/generic_dgram_socket_send_recv_test.go
@@ -0,0 +1,786 @@
+// Copyright 2021 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package generic_dgram_socket_send_recv_test
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"net"
+	"testing"
+	"time"
+
+	"github.com/google/go-cmp/cmp"
+	"golang.org/x/sys/unix"
+	"gvisor.dev/gvisor/pkg/tcpip"
+	"gvisor.dev/gvisor/pkg/tcpip/header"
+	"gvisor.dev/gvisor/test/packetimpact/testbench"
+)
+
+const (
+	// Even though sockets allow larger datagrams we don't test it here as they
+	// need to be fragmented and written out as individual frames.
+
+	maxICMPv4PayloadSize = header.IPv4MinimumMTU - header.EthernetMinimumSize - header.IPv4MinimumSize - header.ICMPv4MinimumSize
+	maxICMPv6PayloadSize = header.IPv6MinimumMTU - header.EthernetMinimumSize - header.IPv6MinimumSize - header.ICMPv6MinimumSize
+	maxUDPv4PayloadSize  = header.IPv4MinimumMTU - header.EthernetMinimumSize - header.IPv4MinimumSize - header.UDPMinimumSize
+	maxUDPv6PayloadSize  = header.IPv6MinimumMTU - header.EthernetMinimumSize - header.IPv6MinimumSize - header.UDPMinimumSize
+)
+
+func maxUDPPayloadSize(addr net.IP) int {
+	if addr.To4() != nil {
+		return maxUDPv4PayloadSize
+	}
+	return maxUDPv6PayloadSize
+}
+
+func init() {
+	testbench.Initialize(flag.CommandLine)
+	testbench.RPCTimeout = 500 * time.Millisecond
+}
+
+func expectedEthLayer(t *testing.T, dut testbench.DUT, socketFD int32, sendTo net.IP) testbench.Layer {
+	t.Helper()
+	dst := func() tcpip.LinkAddress {
+		if isBroadcast(dut, sendTo) {
+			dut.SetSockOptInt(t, socketFD, unix.SOL_SOCKET, unix.SO_BROADCAST, 1)
+
+			// When sending to broadcast (subnet or limited), the expected ethernet
+			// address is also broadcast.
+			return header.EthernetBroadcastAddress
+		}
+		if sendTo.IsMulticast() {
+			if sendTo4 := sendTo.To4(); sendTo4 != nil {
+				return header.EthernetAddressFromMulticastIPv4Address(tcpip.Address(sendTo4))
+			}
+			return header.EthernetAddressFromMulticastIPv6Address(tcpip.Address(sendTo.To16()))
+		}
+		return ""
+	}()
+	var ether testbench.Ether
+	if len(dst) != 0 {
+		ether.DstAddr = &dst
+	}
+	return &ether
+}
+
+type protocolTest interface {
+	Name() string
+	Send(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool)
+	Receive(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool)
+}
+
+func TestSocket(t *testing.T) {
+	dut := testbench.NewDUT(t)
+	subnetBroadcast := dut.Net.SubnetBroadcast()
+
+	for _, proto := range []protocolTest{
+		&icmpV4Test{},
+		&icmpV6Test{},
+		&udpTest{},
+	} {
+		t.Run(proto.Name(), func(t *testing.T) {
+			// Test every combination of bound/unbound, broadcast/multicast/unicast
+			// bound/destination address, and bound/not-bound to device.
+			for _, bindTo := range []net.IP{
+				nil, // Do not bind.
+				net.IPv4zero,
+				net.IPv4bcast,
+				net.IPv4allsys,
+				net.IPv6zero,
+				subnetBroadcast,
+				dut.Net.RemoteIPv4,
+				dut.Net.RemoteIPv6,
+			} {
+				t.Run(fmt.Sprintf("bindTo=%s", bindTo), func(t *testing.T) {
+					for _, sendTo := range []net.IP{
+						net.IPv4bcast,
+						net.IPv4allsys,
+						subnetBroadcast,
+						dut.Net.LocalIPv4,
+						dut.Net.LocalIPv6,
+						dut.Net.RemoteIPv4,
+						dut.Net.RemoteIPv6,
+					} {
+						t.Run(fmt.Sprintf("sendTo=%s", sendTo), func(t *testing.T) {
+							for _, bindToDevice := range []bool{true, false} {
+								t.Run(fmt.Sprintf("bindToDevice=%t", bindToDevice), func(t *testing.T) {
+									t.Run("Send", func(t *testing.T) {
+										proto.Send(t, dut, bindTo, sendTo, bindToDevice)
+									})
+									t.Run("Receive", func(t *testing.T) {
+										proto.Receive(t, dut, bindTo, sendTo, bindToDevice)
+									})
+								})
+							}
+						})
+					}
+				})
+			}
+		})
+	}
+}
+
+type icmpV4TestEnv struct {
+	socketFD int32
+	ident    uint16
+	conn     testbench.IPv4Conn
+	layers   testbench.Layers
+}
+
+type icmpV4Test struct{}
+
+func (test *icmpV4Test) setup(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) icmpV4TestEnv {
+	t.Helper()
+
+	// Tell the DUT to create a socket.
+	var socketFD int32
+	var ident uint16
+
+	if bindTo != nil {
+		socketFD, ident = dut.CreateBoundSocket(t, unix.SOCK_DGRAM, unix.IPPROTO_ICMP, bindTo)
+	} else {
+		// An unbound socket will auto-bind to INADDR_ANY.
+		socketFD = dut.Socket(t, unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_ICMP)
+	}
+	t.Cleanup(func() {
+		dut.Close(t, socketFD)
+	})
+
+	if bindToDevice {
+		dut.SetSockOpt(t, socketFD, unix.SOL_SOCKET, unix.SO_BINDTODEVICE, []byte(dut.Net.RemoteDevName))
+	}
+
+	// Create a socket on the test runner.
+	conn := dut.Net.NewIPv4Conn(t, testbench.IPv4{}, testbench.IPv4{})
+	t.Cleanup(func() {
+		conn.Close(t)
+	})
+
+	return icmpV4TestEnv{
+		socketFD: socketFD,
+		ident:    ident,
+		conn:     conn,
+		layers: testbench.Layers{
+			expectedEthLayer(t, dut, socketFD, sendTo),
+			&testbench.IPv4{
+				DstAddr: testbench.Address(tcpip.Address(sendTo.To4())),
+			},
+		},
+	}
+}
+
+var _ protocolTest = (*icmpV4Test)(nil)
+
+func (*icmpV4Test) Name() string { return "icmpv4" }
+
+func (test *icmpV4Test) Send(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) {
+	if bindTo.To4() == nil || isBroadcastOrMulticast(dut, bindTo) {
+		// ICMPv4 sockets cannot bind to IPv6, broadcast, or multicast
+		// addresses.
+		return
+	}
+
+	isV4 := sendTo.To4() != nil
+
+	// TODO(gvisor.dev/issue/5681): Remove this case once ICMP sockets allow
+	// sending to broadcast and multicast addresses.
+	if (dut.Uname.IsGvisor() || dut.Uname.IsFuchsia()) && isV4 && isBroadcastOrMulticast(dut, sendTo) {
+		// expectPacket cannot be false. In some cases the packet will send, but
+		// with IPv4 destination incorrectly set to RemoteIPv4. It's all bad and
+		// not worth the effort to create a special case when this occurs.
+		t.Skip("TODO(gvisor.dev/issue/5681): Allow sending to broadcast and multicast addresses with ICMP sockets.")
+	}
+
+	expectPacket := isV4 && !sendTo.Equal(dut.Net.RemoteIPv4)
+	switch {
+	case bindTo.Equal(dut.Net.RemoteIPv4):
+		// If we're explicitly bound to an interface's unicast address,
+		// packets are always sent on that interface.
+	case bindToDevice:
+		// If we're explicitly bound to an interface, packets are always
+		// sent on that interface.
+	case !sendTo.Equal(net.IPv4bcast) && !sendTo.IsMulticast():
+		// If we're not sending to limited broadcast or multicast, the route
+		// table will be consulted and packets will be sent on the correct
+		// interface.
+	default:
+		expectPacket = false
+	}
+
+	env := test.setup(t, dut, bindTo, sendTo, bindToDevice)
+
+	for name, payload := range map[string][]byte{
+		"empty":    nil,
+		"small":    []byte("hello world"),
+		"random1k": testbench.GenerateRandomPayload(t, maxICMPv4PayloadSize),
+	} {
+		t.Run(name, func(t *testing.T) {
+			icmpLayer := &testbench.ICMPv4{
+				Type:    testbench.ICMPv4Type(header.ICMPv4Echo),
+				Payload: payload,
+			}
+			bytes, err := icmpLayer.ToBytes()
+			if err != nil {
+				t.Fatalf("icmpLayer.ToBytes() = %s", err)
+			}
+			destSockaddr := unix.SockaddrInet4{}
+			copy(destSockaddr.Addr[:], sendTo.To4())
+
+			// Tell the DUT to send a packet out the ICMP socket.
+			if got, want := dut.SendTo(t, env.socketFD, bytes, 0, &destSockaddr), len(bytes); int(got) != want {
+				t.Fatalf("got dut.SendTo = %d, want %d", got, want)
+			}
+
+			// Verify the test runner received an ICMP packet with the correctly
+			// set "ident".
+			if env.ident != 0 {
+				icmpLayer.Ident = &env.ident
+			}
+			want := append(env.layers, icmpLayer)
+			if got, ok := env.conn.ListenForFrame(t, want, time.Second); !ok && expectPacket {
+				t.Fatalf("did not receive expected frame matching %s\nGot frames: %s", want, got)
+			} else if ok && !expectPacket {
+				matchedFrame := got[len(got)-1]
+				t.Fatalf("got unexpected frame matching %s\nGot frame: %s", want, matchedFrame)
+			}
+		})
+	}
+}
+
+func (test *icmpV4Test) Receive(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) {
+	if bindTo.To4() == nil || isBroadcastOrMulticast(dut, bindTo) {
+		// ICMPv4 sockets cannot bind to IPv6, broadcast, or multicast
+		// addresses.
+		return
+	}
+
+	expectPacket := (bindTo.Equal(dut.Net.RemoteIPv4) || bindTo.Equal(net.IPv4zero)) && sendTo.Equal(dut.Net.RemoteIPv4)
+
+	// TODO(gvisor.dev/issue/5763): Remove this if statement once gVisor
+	// restricts ICMP sockets to receive only from unicast addresses.
+	if (dut.Uname.IsGvisor() || dut.Uname.IsFuchsia()) && bindTo.Equal(net.IPv4zero) && isBroadcastOrMulticast(dut, sendTo) {
+		expectPacket = true
+	}
+
+	env := test.setup(t, dut, bindTo, sendTo, bindToDevice)
+
+	for name, payload := range map[string][]byte{
+		"empty":    nil,
+		"small":    []byte("hello world"),
+		"random1k": testbench.GenerateRandomPayload(t, maxICMPv4PayloadSize),
+	} {
+		t.Run(name, func(t *testing.T) {
+			icmpLayer := &testbench.ICMPv4{
+				Type:    testbench.ICMPv4Type(header.ICMPv4EchoReply),
+				Payload: payload,
+			}
+			if env.ident != 0 {
+				icmpLayer.Ident = &env.ident
+			}
+
+			// Send an ICMPv4 packet from the test runner to the DUT.
+			frame := env.conn.CreateFrame(t, env.layers, icmpLayer)
+			env.conn.SendFrame(t, frame)
+
+			// Verify the behavior of the ICMP socket on the DUT.
+			if expectPacket {
+				payload, err := icmpLayer.ToBytes()
+				if err != nil {
+					t.Fatalf("icmpLayer.ToBytes() = %s", err)
+				}
+
+				// Receive one extra byte to assert the length of the
+				// packet received in the case where the packet contains
+				// more data than expected.
+				len := int32(len(payload)) + 1
+				got, want := dut.Recv(t, env.socketFD, len, 0), payload
+				if diff := cmp.Diff(want, got); diff != "" {
+					t.Errorf("received payload does not match sent payload, diff (-want, +got):\n%s", diff)
+				}
+			} else {
+				// Expected receive error, set a short receive timeout.
+				dut.SetSockOptTimeval(
+					t,
+					env.socketFD,
+					unix.SOL_SOCKET,
+					unix.SO_RCVTIMEO,
+					&unix.Timeval{
+						Sec:  1,
+						Usec: 0,
+					},
+				)
+				ret, recvPayload, errno := dut.RecvWithErrno(context.Background(), t, env.socketFD, maxICMPv4PayloadSize, 0)
+				if errno != unix.EAGAIN || errno != unix.EWOULDBLOCK {
+					t.Errorf("Recv got unexpected result, ret=%d, payload=%q, errno=%s", ret, recvPayload, errno)
+				}
+			}
+		})
+	}
+}
+
+type icmpV6TestEnv struct {
+	socketFD int32
+	ident    uint16
+	conn     testbench.IPv6Conn
+	layers   testbench.Layers
+}
+
+// icmpV6Test and icmpV4Test look substantially similar at first look, but have
+// enough subtle differences in setup and test expectations to discourage
+// refactoring:
+//  - Different IP layers
+//  - Different testbench.Connections
+//  - Different UNIX domain and proto arguments
+//  - Different expectPacket and wantErrno for send and receive
+type icmpV6Test struct{}
+
+func (test *icmpV6Test) setup(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) icmpV6TestEnv {
+	t.Helper()
+
+	// Tell the DUT to create a socket.
+	var socketFD int32
+	var ident uint16
+
+	if bindTo != nil {
+		socketFD, ident = dut.CreateBoundSocket(t, unix.SOCK_DGRAM, unix.IPPROTO_ICMPV6, bindTo)
+	} else {
+		// An unbound socket will auto-bind to IN6ADDR_ANY_INIT.
+		socketFD = dut.Socket(t, unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_ICMPV6)
+	}
+	t.Cleanup(func() {
+		dut.Close(t, socketFD)
+	})
+
+	if bindToDevice {
+		dut.SetSockOpt(t, socketFD, unix.SOL_SOCKET, unix.SO_BINDTODEVICE, []byte(dut.Net.RemoteDevName))
+	}
+
+	// Create a socket on the test runner.
+	conn := dut.Net.NewIPv6Conn(t, testbench.IPv6{}, testbench.IPv6{})
+	t.Cleanup(func() {
+		conn.Close(t)
+	})
+
+	return icmpV6TestEnv{
+		socketFD: socketFD,
+		ident:    ident,
+		conn:     conn,
+		layers: testbench.Layers{
+			expectedEthLayer(t, dut, socketFD, sendTo),
+			&testbench.IPv6{
+				DstAddr: testbench.Address(tcpip.Address(sendTo.To16())),
+			},
+		},
+	}
+}
+
+var _ protocolTest = (*icmpV6Test)(nil)
+
+func (*icmpV6Test) Name() string { return "icmpv6" }
+
+func (test *icmpV6Test) Send(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) {
+	if bindTo.To4() != nil || bindTo.IsMulticast() {
+		// ICMPv6 sockets cannot bind to IPv4 or multicast addresses.
+		return
+	}
+
+	expectPacket := sendTo.Equal(dut.Net.LocalIPv6)
+	wantErrno := unix.Errno(0)
+
+	if sendTo.To4() != nil {
+		wantErrno = unix.EINVAL
+	}
+
+	// TODO(gvisor.dev/issue/5966): Remove this if statement once ICMPv6 sockets
+	// return EINVAL after calling sendto with an IPv4 address.
+	if (dut.Uname.IsGvisor() || dut.Uname.IsFuchsia()) && sendTo.To4() != nil {
+		switch {
+		case bindTo.Equal(dut.Net.RemoteIPv6):
+			wantErrno = unix.ENETUNREACH
+		case bindTo.Equal(net.IPv6zero) || bindTo == nil:
+			wantErrno = unix.Errno(0)
+		}
+	}
+
+	env := test.setup(t, dut, bindTo, sendTo, bindToDevice)
+
+	for name, payload := range map[string][]byte{
+		"empty":    nil,
+		"small":    []byte("hello world"),
+		"random1k": testbench.GenerateRandomPayload(t, maxICMPv6PayloadSize),
+	} {
+		t.Run(name, func(t *testing.T) {
+			icmpLayer := &testbench.ICMPv6{
+				Type:    testbench.ICMPv6Type(header.ICMPv6EchoRequest),
+				Payload: payload,
+			}
+			bytes, err := icmpLayer.ToBytes()
+			if err != nil {
+				t.Fatalf("icmpLayer.ToBytes() = %s", err)
+			}
+			destSockaddr := unix.SockaddrInet6{
+				ZoneId: dut.Net.RemoteDevID,
+			}
+			copy(destSockaddr.Addr[:], sendTo.To16())
+
+			// Tell the DUT to send a packet out the ICMPv6 socket.
+			ctx, cancel := context.WithTimeout(context.Background(), testbench.RPCTimeout)
+			defer cancel()
+			gotRet, gotErrno := dut.SendToWithErrno(ctx, t, env.socketFD, bytes, 0, &destSockaddr)
+
+			if gotErrno != wantErrno {
+				t.Fatalf("got dut.SendToWithErrno(_, _, %d, _, _, %s) = (_, %s), want = (_, %s)", env.socketFD, sendTo, gotErrno, wantErrno)
+			}
+			if wantErrno != 0 {
+				return
+			}
+			if got, want := int(gotRet), len(bytes); got != want {
+				t.Fatalf("got dut.SendToWithErrno(_, _, %d, _, _, %s) = (%d, _), want = (%d, _)", env.socketFD, sendTo, got, want)
+			}
+
+			// Verify the test runner received an ICMPv6 packet with the
+			// correctly set "ident".
+			if env.ident != 0 {
+				icmpLayer.Ident = &env.ident
+			}
+			want := append(env.layers, icmpLayer)
+			if got, ok := env.conn.ListenForFrame(t, want, time.Second); !ok && expectPacket {
+				t.Fatalf("did not receive expected frame matching %s\nGot frames: %s", want, got)
+			} else if ok && !expectPacket {
+				matchedFrame := got[len(got)-1]
+				t.Fatalf("got unexpected frame matching %s\nGot frame: %s", want, matchedFrame)
+			}
+		})
+	}
+}
+
+func (test *icmpV6Test) Receive(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) {
+	if bindTo.To4() != nil || bindTo.IsMulticast() {
+		// ICMPv6 sockets cannot bind to IPv4 or multicast addresses.
+		return
+	}
+
+	expectPacket := true
+	switch {
+	case bindTo.Equal(dut.Net.RemoteIPv6) && sendTo.Equal(dut.Net.RemoteIPv6):
+	case bindTo.Equal(net.IPv6zero) && sendTo.Equal(dut.Net.RemoteIPv6):
+	case bindTo.Equal(net.IPv6zero) && sendTo.Equal(net.IPv6linklocalallnodes):
+	default:
+		expectPacket = false
+	}
+
+	// TODO(gvisor.dev/issue/5763): Remove this if statement once gVisor
+	// restricts ICMP sockets to receive only from unicast addresses.
+	if (dut.Uname.IsGvisor() || dut.Uname.IsFuchsia()) && bindTo.Equal(net.IPv6zero) && isBroadcastOrMulticast(dut, sendTo) {
+		expectPacket = false
+	}
+
+	env := test.setup(t, dut, bindTo, sendTo, bindToDevice)
+
+	for name, payload := range map[string][]byte{
+		"empty":    nil,
+		"small":    []byte("hello world"),
+		"random1k": testbench.GenerateRandomPayload(t, maxICMPv6PayloadSize),
+	} {
+		t.Run(name, func(t *testing.T) {
+			icmpLayer := &testbench.ICMPv6{
+				Type:    testbench.ICMPv6Type(header.ICMPv6EchoReply),
+				Payload: payload,
+			}
+			if env.ident != 0 {
+				icmpLayer.Ident = &env.ident
+			}
+
+			// Send an ICMPv6 packet from the test runner to the DUT.
+			frame := env.conn.CreateFrame(t, env.layers, icmpLayer)
+			env.conn.SendFrame(t, frame)
+
+			// Verify the behavior of the ICMPv6 socket on the DUT.
+			if expectPacket {
+				payload, err := icmpLayer.ToBytes()
+				if err != nil {
+					t.Fatalf("icmpLayer.ToBytes() = %s", err)
+				}
+
+				// Receive one extra byte to assert the length of the
+				// packet received in the case where the packet contains
+				// more data than expected.
+				len := int32(len(payload)) + 1
+				got, want := dut.Recv(t, env.socketFD, len, 0), payload
+				if diff := cmp.Diff(want, got); diff != "" {
+					t.Errorf("received payload does not match sent payload, diff (-want, +got):\n%s", diff)
+				}
+			} else {
+				// Expected receive error, set a short receive timeout.
+				dut.SetSockOptTimeval(
+					t,
+					env.socketFD,
+					unix.SOL_SOCKET,
+					unix.SO_RCVTIMEO,
+					&unix.Timeval{
+						Sec:  1,
+						Usec: 0,
+					},
+				)
+				ret, recvPayload, errno := dut.RecvWithErrno(context.Background(), t, env.socketFD, maxICMPv6PayloadSize, 0)
+				if errno != unix.EAGAIN || errno != unix.EWOULDBLOCK {
+					t.Errorf("Recv got unexpected result, ret=%d, payload=%q, errno=%s", ret, recvPayload, errno)
+				}
+			}
+		})
+	}
+}
+
+type udpConn interface {
+	SrcPort(*testing.T) uint16
+	SendFrame(*testing.T, testbench.Layers, ...testbench.Layer)
+	ListenForFrame(*testing.T, testbench.Layers, time.Duration) ([]testbench.Layers, bool)
+	Close(*testing.T)
+}
+
+type udpTestEnv struct {
+	socketFD int32
+	conn     udpConn
+	layers   testbench.Layers
+}
+
+type udpTest struct{}
+
+func (test *udpTest) setup(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) udpTestEnv {
+	t.Helper()
+
+	var (
+		socketFD                 int32
+		outgoingUDP, incomingUDP testbench.UDP
+	)
+
+	// Tell the DUT to create a socket.
+	if bindTo != nil {
+		var remotePort uint16
+		socketFD, remotePort = dut.CreateBoundSocket(t, unix.SOCK_DGRAM, unix.IPPROTO_UDP, bindTo)
+		outgoingUDP.DstPort = &remotePort
+		incomingUDP.SrcPort = &remotePort
+	} else {
+		// An unbound socket will auto-bind to INADDR_ANY.
+		socketFD = dut.Socket(t, unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_UDP)
+	}
+	t.Cleanup(func() {
+		dut.Close(t, socketFD)
+	})
+
+	if bindToDevice {
+		dut.SetSockOpt(t, socketFD, unix.SOL_SOCKET, unix.SO_BINDTODEVICE, []byte(dut.Net.RemoteDevName))
+	}
+
+	// Create a socket on the test runner.
+	var conn udpConn
+	var ipLayer testbench.Layer
+	if addr := sendTo.To4(); addr != nil {
+		udpConn := dut.Net.NewUDPIPv4(t, outgoingUDP, incomingUDP)
+		conn = &udpConn
+		ipLayer = &testbench.IPv4{
+			DstAddr: testbench.Address(tcpip.Address(addr)),
+		}
+	} else {
+		udpConn := dut.Net.NewUDPIPv6(t, outgoingUDP, incomingUDP)
+		conn = &udpConn
+		ipLayer = &testbench.IPv6{
+			DstAddr: testbench.Address(tcpip.Address(sendTo.To16())),
+		}
+	}
+	t.Cleanup(func() {
+		conn.Close(t)
+	})
+
+	return udpTestEnv{
+		socketFD: socketFD,
+		conn:     conn,
+		layers: testbench.Layers{
+			expectedEthLayer(t, dut, socketFD, sendTo),
+			ipLayer,
+			&incomingUDP,
+		},
+	}
+}
+
+var _ protocolTest = (*udpTest)(nil)
+
+func (*udpTest) Name() string { return "udp" }
+
+func (test *udpTest) Send(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) {
+	canSend := bindTo == nil || bindTo.Equal(net.IPv6zero) || sameIPVersion(sendTo, bindTo)
+	expectPacket := canSend && !isRemoteAddr(dut, sendTo)
+	switch {
+	case bindTo.Equal(dut.Net.RemoteIPv4):
+		// If we're explicitly bound to an interface's unicast address,
+		// packets are always sent on that interface.
+	case bindToDevice:
+		// If we're explicitly bound to an interface, packets are always
+		// sent on that interface.
+	case !sendTo.Equal(net.IPv4bcast) && !sendTo.IsMulticast():
+		// If we're not sending to limited broadcast, multicast, or local, the
+		// route table will be consulted and packets will be sent on the correct
+		// interface.
+	default:
+		expectPacket = false
+	}
+
+	wantErrno := unix.Errno(0)
+	switch {
+	case !canSend && bindTo.To4() != nil:
+		wantErrno = unix.EAFNOSUPPORT
+	case !canSend && bindTo.To4() == nil:
+		wantErrno = unix.ENETUNREACH
+	}
+
+	// TODO(gvisor.dev/issue/5967): Remove this if statement once UDPv4 sockets
+	// returns EAFNOSUPPORT after calling sendto with an IPv6 address.
+	if dut.Uname.IsGvisor() && !canSend && bindTo.To4() != nil {
+		wantErrno = unix.EINVAL
+	}
+
+	env := test.setup(t, dut, bindTo, sendTo, bindToDevice)
+
+	for name, payload := range map[string][]byte{
+		"empty":    nil,
+		"small":    []byte("hello world"),
+		"random1k": testbench.GenerateRandomPayload(t, maxUDPPayloadSize(bindTo)),
+	} {
+		t.Run(name, func(t *testing.T) {
+			var destSockaddr unix.Sockaddr
+			if sendTo4 := sendTo.To4(); sendTo4 != nil {
+				addr := unix.SockaddrInet4{
+					Port: int(env.conn.SrcPort(t)),
+				}
+				copy(addr.Addr[:], sendTo4)
+				destSockaddr = &addr
+			} else {
+				addr := unix.SockaddrInet6{
+					Port:   int(env.conn.SrcPort(t)),
+					ZoneId: dut.Net.RemoteDevID,
+				}
+				copy(addr.Addr[:], sendTo.To16())
+				destSockaddr = &addr
+			}
+
+			// Tell the DUT to send a packet out the UDP socket.
+			ctx, cancel := context.WithTimeout(context.Background(), testbench.RPCTimeout)
+			defer cancel()
+			gotRet, gotErrno := dut.SendToWithErrno(ctx, t, env.socketFD, payload, 0, destSockaddr)
+
+			if gotErrno != wantErrno {
+				t.Fatalf("got dut.SendToWithErrno(_, _, %d, _, _, %s) = (_, %s), want = (_, %s)", env.socketFD, sendTo, gotErrno, wantErrno)
+			}
+			if wantErrno != 0 {
+				return
+			}
+			if got, want := int(gotRet), len(payload); got != want {
+				t.Fatalf("got dut.SendToWithErrno(_, _, %d, _, _, %s) = (%d, _), want = (%d, _)", env.socketFD, sendTo, got, want)
+			}
+
+			// Verify the test runner received a UDP packet with the
+			// correct payload.
+			want := append(env.layers, &testbench.Payload{
+				Bytes: payload,
+			})
+			if got, ok := env.conn.ListenForFrame(t, want, time.Second); !ok && expectPacket {
+				t.Fatalf("did not receive expected frame matching %s\nGot frames: %s", want, got)
+			} else if ok && !expectPacket {
+				matchedFrame := got[len(got)-1]
+				t.Fatalf("got unexpected frame matching %s\nGot frame: %s", want, matchedFrame)
+			}
+		})
+	}
+}
+
+func (test *udpTest) Receive(t *testing.T, dut testbench.DUT, bindTo, sendTo net.IP, bindToDevice bool) {
+	subnetBroadcast := dut.Net.SubnetBroadcast()
+
+	expectPacket := true
+	switch {
+	case bindTo.Equal(sendTo):
+	case bindTo.Equal(net.IPv4zero) && sameIPVersion(bindTo, sendTo) && !sendTo.Equal(dut.Net.LocalIPv4):
+	case bindTo.Equal(net.IPv6zero) && isBroadcast(dut, sendTo):
+	case bindTo.Equal(net.IPv6zero) && isRemoteAddr(dut, sendTo):
+	case bindTo.Equal(subnetBroadcast) && sendTo.Equal(subnetBroadcast):
+	default:
+		expectPacket = false
+	}
+
+	// TODO(gvisor.dev/issue/5956): Remove this if statement once gVisor
+	// restricts ICMP sockets to receive only from unicast addresses.
+	if (dut.Uname.IsGvisor() || dut.Uname.IsFuchsia()) && bindTo.Equal(net.IPv6zero) && sendTo.Equal(net.IPv4allsys) {
+		expectPacket = true
+	}
+
+	env := test.setup(t, dut, bindTo, sendTo, bindToDevice)
+	maxPayloadSize := maxUDPPayloadSize(bindTo)
+
+	for name, payload := range map[string][]byte{
+		"empty":    nil,
+		"small":    []byte("hello world"),
+		"random1k": testbench.GenerateRandomPayload(t, maxPayloadSize),
+	} {
+		t.Run(name, func(t *testing.T) {
+			// Send a UDP packet from the test runner to the DUT.
+			env.conn.SendFrame(t, env.layers, &testbench.Payload{Bytes: payload})
+
+			// Verify the behavior of the ICMP socket on the DUT.
+			if expectPacket {
+				// Receive one extra byte to assert the length of the
+				// packet received in the case where the packet contains
+				// more data than expected.
+				len := int32(len(payload)) + 1
+				got, want := dut.Recv(t, env.socketFD, len, 0), payload
+				if diff := cmp.Diff(want, got); diff != "" {
+					t.Errorf("received payload does not match sent payload, diff (-want, +got):\n%s", diff)
+				}
+			} else {
+				// Expected receive error, set a short receive timeout.
+				dut.SetSockOptTimeval(
+					t,
+					env.socketFD,
+					unix.SOL_SOCKET,
+					unix.SO_RCVTIMEO,
+					&unix.Timeval{
+						Sec:  1,
+						Usec: 0,
+					},
+				)
+				ret, recvPayload, errno := dut.RecvWithErrno(context.Background(), t, env.socketFD, int32(maxPayloadSize), 0)
+				if errno != unix.EAGAIN || errno != unix.EWOULDBLOCK {
+					t.Errorf("Recv got unexpected result, ret=%d, payload=%q, errno=%s", ret, recvPayload, errno)
+				}
+			}
+		})
+	}
+}
+
+func isBroadcast(dut testbench.DUT, ip net.IP) bool {
+	return ip.Equal(net.IPv4bcast) || ip.Equal(dut.Net.SubnetBroadcast())
+}
+
+func isBroadcastOrMulticast(dut testbench.DUT, ip net.IP) bool {
+	return isBroadcast(dut, ip) || ip.IsMulticast()
+}
+
+func sameIPVersion(a, b net.IP) bool {
+	return (a.To4() == nil) == (b.To4() == nil)
+}
+
+func isRemoteAddr(dut testbench.DUT, ip net.IP) bool {
+	return ip.Equal(dut.Net.RemoteIPv4) || ip.Equal(dut.Net.RemoteIPv6)
+}
diff --git a/test/packetimpact/tests/udp_send_recv_dgram_test.go b/test/packetimpact/tests/udp_send_recv_dgram_test.go
deleted file mode 100644
index 230b012c7..000000000
--- a/test/packetimpact/tests/udp_send_recv_dgram_test.go
+++ /dev/null
@@ -1,329 +0,0 @@
-// Copyright 2020 The gVisor Authors.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//     http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-
-package udp_send_recv_dgram_test
-
-import (
-	"context"
-	"flag"
-	"fmt"
-	"net"
-	"testing"
-	"time"
-
-	"github.com/google/go-cmp/cmp"
-	"golang.org/x/sys/unix"
-	"gvisor.dev/gvisor/pkg/tcpip"
-	"gvisor.dev/gvisor/pkg/tcpip/header"
-	"gvisor.dev/gvisor/test/packetimpact/testbench"
-)
-
-func init() {
-	testbench.Initialize(flag.CommandLine)
-	testbench.RPCTimeout = 500 * time.Millisecond
-}
-
-type udpConn interface {
-	SrcPort(*testing.T) uint16
-	SendFrame(*testing.T, testbench.Layers, ...testbench.Layer)
-	ExpectFrame(*testing.T, testbench.Layers, time.Duration) (testbench.Layers, error)
-	Close(*testing.T)
-}
-
-type testCase struct {
-	bindTo, sendTo                            net.IP
-	sendToBroadcast, bindToDevice, expectData bool
-}
-
-func TestUDP(t *testing.T) {
-	dut := testbench.NewDUT(t)
-	subnetBcast := func() net.IP {
-		subnet := (&tcpip.AddressWithPrefix{
-			Address:   tcpip.Address(dut.Net.RemoteIPv4.To4()),
-			PrefixLen: dut.Net.IPv4PrefixLength,
-		}).Subnet()
-		return net.IP(subnet.Broadcast())
-	}()
-
-	t.Run("Send", func(t *testing.T) {
-		var testCases []testCase
-		// Test every valid combination of bound/unbound, broadcast/multicast/unicast
-		// bound/destination address, and bound/not-bound to device.
-		for _, bindTo := range []net.IP{
-			nil, // Do not bind.
-			net.IPv4zero,
-			net.IPv4bcast,
-			net.IPv4allsys,
-			subnetBcast,
-			dut.Net.RemoteIPv4,
-			dut.Net.RemoteIPv6,
-		} {
-			for _, sendTo := range []net.IP{
-				net.IPv4bcast,
-				net.IPv4allsys,
-				subnetBcast,
-				dut.Net.LocalIPv4,
-				dut.Net.LocalIPv6,
-			} {
-				// Cannot send to an IPv4 address from a socket bound to IPv6 (except for IPv4-mapped IPv6),
-				// and viceversa.
-				if bindTo != nil && ((bindTo.To4() == nil) != (sendTo.To4() == nil)) {
-					continue
-				}
-				for _, bindToDevice := range []bool{true, false} {
-					expectData := true
-					switch {
-					case bindTo.Equal(dut.Net.RemoteIPv4):
-						// If we're explicitly bound to an interface's unicast address,
-						// packets are always sent on that interface.
-					case bindToDevice:
-						// If we're explicitly bound to an interface, packets are always
-						// sent on that interface.
-					case !sendTo.Equal(net.IPv4bcast) && !sendTo.IsMulticast():
-						// If we're not sending to limited broadcast or multicast, the route table
-						// will be consulted and packets will be sent on the correct interface.
-					default:
-						expectData = false
-					}
-					testCases = append(
-						testCases,
-						testCase{
-							bindTo:          bindTo,
-							sendTo:          sendTo,
-							sendToBroadcast: sendTo.Equal(subnetBcast) || sendTo.Equal(net.IPv4bcast),
-							bindToDevice:    bindToDevice,
-							expectData:      expectData,
-						},
-					)
-				}
-			}
-		}
-		for _, tc := range testCases {
-			boundTestCaseName := "unbound"
-			if tc.bindTo != nil {
-				boundTestCaseName = fmt.Sprintf("bindTo=%s", tc.bindTo)
-			}
-			t.Run(fmt.Sprintf("%s/sendTo=%s/bindToDevice=%t/expectData=%t", boundTestCaseName, tc.sendTo, tc.bindToDevice, tc.expectData), func(t *testing.T) {
-				runTestCase(
-					t,
-					dut,
-					tc,
-					func(t *testing.T, dut testbench.DUT, conn udpConn, socketFD int32, tc testCase, payload []byte, layers testbench.Layers) {
-						var destSockaddr unix.Sockaddr
-						if sendTo4 := tc.sendTo.To4(); sendTo4 != nil {
-							addr := unix.SockaddrInet4{
-								Port: int(conn.SrcPort(t)),
-							}
-							copy(addr.Addr[:], sendTo4)
-							destSockaddr = &addr
-						} else {
-							addr := unix.SockaddrInet6{
-								Port:   int(conn.SrcPort(t)),
-								ZoneId: dut.Net.RemoteDevID,
-							}
-							copy(addr.Addr[:], tc.sendTo.To16())
-							destSockaddr = &addr
-						}
-						if got, want := dut.SendTo(t, socketFD, payload, 0, destSockaddr), len(payload); int(got) != want {
-							t.Fatalf("got dut.SendTo = %d, want %d", got, want)
-						}
-						layers = append(layers, &testbench.Payload{
-							Bytes: payload,
-						})
-						_, err := conn.ExpectFrame(t, layers, time.Second)
-
-						if !tc.expectData && err == nil {
-							t.Fatal("received unexpected packet, socket is not bound to device")
-						}
-						if err != nil && tc.expectData {
-							t.Fatal(err)
-						}
-					},
-				)
-			})
-		}
-	})
-	t.Run("Recv", func(t *testing.T) {
-		// Test every valid combination of broadcast/multicast/unicast
-		// bound/destination address, and bound/not-bound to device.
-		var testCases []testCase
-		for _, addr := range []net.IP{
-			net.IPv4bcast,
-			net.IPv4allsys,
-			dut.Net.RemoteIPv4,
-			dut.Net.RemoteIPv6,
-		} {
-			for _, bindToDevice := range []bool{true, false} {
-				testCases = append(
-					testCases,
-					testCase{
-						bindTo:          addr,
-						sendTo:          addr,
-						sendToBroadcast: addr.Equal(subnetBcast) || addr.Equal(net.IPv4bcast),
-						bindToDevice:    bindToDevice,
-						expectData:      true,
-					},
-				)
-			}
-		}
-		for _, bindTo := range []net.IP{
-			net.IPv4zero,
-			subnetBcast,
-			dut.Net.RemoteIPv4,
-		} {
-			for _, sendTo := range []net.IP{
-				subnetBcast,
-				net.IPv4bcast,
-				net.IPv4allsys,
-			} {
-				// TODO(gvisor.dev/issue/4896): Add bindTo=subnetBcast/sendTo=IPv4bcast
-				// and bindTo=subnetBcast/sendTo=IPv4allsys test cases.
-				if bindTo.Equal(subnetBcast) && (sendTo.Equal(net.IPv4bcast) || sendTo.IsMulticast()) {
-					continue
-				}
-				// Expect that a socket bound to a unicast address does not receive
-				// packets sent to an address other than the bound unicast address.
-				//
-				// Note: we cannot use net.IP.IsGlobalUnicast to test this condition
-				// because IsGlobalUnicast does not check whether the address is the
-				// subnet broadcast, and returns true in that case.
-				expectData := !bindTo.Equal(dut.Net.RemoteIPv4) || sendTo.Equal(dut.Net.RemoteIPv4)
-				for _, bindToDevice := range []bool{true, false} {
-					testCases = append(
-						testCases,
-						testCase{
-							bindTo:          bindTo,
-							sendTo:          sendTo,
-							sendToBroadcast: sendTo.Equal(subnetBcast) || sendTo.Equal(net.IPv4bcast),
-							bindToDevice:    bindToDevice,
-							expectData:      expectData,
-						},
-					)
-				}
-			}
-		}
-		for _, tc := range testCases {
-			t.Run(fmt.Sprintf("bindTo=%s/sendTo=%s/bindToDevice=%t/expectData=%t", tc.bindTo, tc.sendTo, tc.bindToDevice, tc.expectData), func(t *testing.T) {
-				runTestCase(
-					t,
-					dut,
-					tc,
-					func(t *testing.T, dut testbench.DUT, conn udpConn, socketFD int32, tc testCase, payload []byte, layers testbench.Layers) {
-						conn.SendFrame(t, layers, &testbench.Payload{Bytes: payload})
-
-						if tc.expectData {
-							got, want := dut.Recv(t, socketFD, int32(len(payload)+1), 0), payload
-							if diff := cmp.Diff(want, got); diff != "" {
-								t.Errorf("received payload does not match sent payload, diff (-want, +got):\n%s", diff)
-							}
-						} else {
-							// Expected receive error, set a short receive timeout.
-							dut.SetSockOptTimeval(
-								t,
-								socketFD,
-								unix.SOL_SOCKET,
-								unix.SO_RCVTIMEO,
-								&unix.Timeval{
-									Sec:  1,
-									Usec: 0,
-								},
-							)
-							ret, recvPayload, errno := dut.RecvWithErrno(context.Background(), t, socketFD, 100, 0)
-							if errno != unix.EAGAIN || errno != unix.EWOULDBLOCK {
-								t.Errorf("Recv got unexpected result, ret=%d, payload=%q, errno=%s", ret, recvPayload, errno)
-							}
-						}
-					},
-				)
-			})
-		}
-	})
-}
-
-func runTestCase(
-	t *testing.T,
-	dut testbench.DUT,
-	tc testCase,
-	runTc func(t *testing.T, dut testbench.DUT, conn udpConn, socketFD int32, tc testCase, payload []byte, layers testbench.Layers),
-) {
-	var (
-		socketFD                 int32
-		outgoingUDP, incomingUDP testbench.UDP
-	)
-	if tc.bindTo != nil {
-		var remotePort uint16
-		socketFD, remotePort = dut.CreateBoundSocket(t, unix.SOCK_DGRAM, unix.IPPROTO_UDP, tc.bindTo)
-		outgoingUDP.DstPort = &remotePort
-		incomingUDP.SrcPort = &remotePort
-	} else {
-		// An unbound socket will auto-bind to INNADDR_ANY and a random
-		// port on sendto.
-		socketFD = dut.Socket(t, unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_UDP)
-	}
-	defer dut.Close(t, socketFD)
-	if tc.bindToDevice {
-		dut.SetSockOpt(t, socketFD, unix.SOL_SOCKET, unix.SO_BINDTODEVICE, []byte(dut.Net.RemoteDevName))
-	}
-
-	var ethernetLayer testbench.Ether
-	if tc.sendToBroadcast {
-		dut.SetSockOptInt(t, socketFD, unix.SOL_SOCKET, unix.SO_BROADCAST, 1)
-
-		// When sending to broadcast (subnet or limited), the expected ethernet
-		// address is also broadcast.
-		ethernetBroadcastAddress := header.EthernetBroadcastAddress
-		ethernetLayer.DstAddr = &ethernetBroadcastAddress
-	} else if tc.sendTo.IsMulticast() {
-		ethernetMulticastAddress := header.EthernetAddressFromMulticastIPv4Address(tcpip.Address(tc.sendTo.To4()))
-		ethernetLayer.DstAddr = &ethernetMulticastAddress
-	}
-	expectedLayers := testbench.Layers{&ethernetLayer}
-
-	var conn udpConn
-	if sendTo4 := tc.sendTo.To4(); sendTo4 != nil {
-		v4Conn := dut.Net.NewUDPIPv4(t, outgoingUDP, incomingUDP)
-		conn = &v4Conn
-		expectedLayers = append(
-			expectedLayers,
-			&testbench.IPv4{
-				DstAddr: testbench.Address(tcpip.Address(sendTo4)),
-			},
-		)
-	} else {
-		v6Conn := dut.Net.NewUDPIPv6(t, outgoingUDP, incomingUDP)
-		conn = &v6Conn
-		expectedLayers = append(
-			expectedLayers,
-			&testbench.IPv6{
-				DstAddr: testbench.Address(tcpip.Address(tc.sendTo)),
-			},
-		)
-	}
-	defer conn.Close(t)
-
-	expectedLayers = append(expectedLayers, &incomingUDP)
-	for _, v := range []struct {
-		name    string
-		payload []byte
-	}{
-		{"emptypayload", nil},
-		{"small payload", []byte("hello world")},
-		{"1kPayload", testbench.GenerateRandomPayload(t, 1<<10)},
-		// Even though UDP allows larger dgrams we don't test it here as
-		// they need to be fragmented and written out as individual
-		// frames.
-	} {
-		runTc(t, dut, conn, socketFD, tc, v.payload, expectedLayers)
-	}
-}